MultiverSeg

Project Page | Paper

Official implementation of MultiverSeg: Scalable Interactive Segmentation of Biomedical Imaging Datasets with In-Context Guidance accepted at ICCV 2025

Hallee E. Wong, Jose Javier Gonzalez Ortiz, John Guttag, Adrian V. Dalca

Updates

• (2025-09-24) released v1 weights used in the ICCV paper
• (2025-08-31) ICCV camera-ready posted on arxiv with additional evaluations on TotalSegmentator and more baselines
• (2025-07-01) 3D Slicer extension developed by Kitware Inc. released: https://github.com/dalcalab/SlicerMultiverSeg
• (2025-06-25) MultiverSeg was accepted to ICCV 2025!
• (2025-01-26) inference code and v0 weights released
• (2024-12-19) arxiv preprint released!

Models

We provide pre-trained weights here.

• v0: initial release (shorter training schedule).

• v1: trained for longer, with improved performance. These are the weights used in the ICCV 2025 paper.

Installation

You can install multiverseg in two ways:

• With pip:

pip install git+https://github.com/halleewong/MultiverSeg.git

• Manually: cloning it and installing dependencies

git clone https://github.com/halleewong/MultiverSeg
python -m pip install -r ./MultiverSeg/requirements.txt
export PYTHONPATH="$PYTHONPATH:$(realpath ./MultiverSeg)"

Getting Started

First download the model checkpoints

cd checkpoints
./download.sh

Then see ./notebooks/inference.ipynb for a tutorial.

Acknowledgements

This project builds extensively on code originally developed for ScribblePrompt and UniverSeg

Citation

If you find our work or any of our materials useful, please cite our paper:

@article{wong2025multiverseg,
  title={MultiverSeg: Scalable Interactive Segmentation of Biomedical Imaging Datasets with In-Context Guidance},
  author={Hallee E. Wong and Jose Javier Gonzalez Ortiz and John Guttag and Adrian V. Dalca},
  journal={International Conference on Computer Vision},
  year={2025},
}